Feature Fusion For Cervical Cancer Detection Using Colposcopic Images

摘要

Cervical cancer kills 260,000 women annually, and nearly 85% of these deaths occur in developing nations. Also it is the leading cause of cancer deaths in women. Disparities of health and poverty play a large role in this high mortality rate. This paper presents an automatic cervical cancer detection technique using colposcopic images. Wavelet and statistical based features are used to distinguish normal and abnormal tissue. The statistical features such as mean, standard deviation and skewness are obtained in the spatial domain. The wavelet energies are extracted from the wavelet decomposed image. Then these features are combined to form the feature vector and used for the detection. The segmented cancer region shows that the proposed fusion approach can detect the cancer affected region accurately than the wavelet and statistical features based approaches.